code-review

Reviews git diffs against a fixed point using parallel correctness, standards, spec, and test reviewers.

3|Updated Jan 10, 2026
One-click install
npx skills add https://github.com/MaxWolf-01/agents --skill code-review-maxwolf-01
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: code-review
Source: https://github.com/MaxWolf-01/agents/tree/main/mx/skills/code-review
Command: npx skills add https://github.com/MaxWolf-01/agents --skill code-review-maxwolf-01

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Reviewing a branch or work-in-progress diff by hand is slow and inconsistent: reviewers miss broken callers, convention drift, spec deviations, and tests that would pass without the change. This Skill structures the review of any diff since a fixed point (commit, branch, tag, or merge-base) into four independent axes checked by parallel subagents, with findings written to disk and disposed as fixed, filed, or declined. ## Core Features & Use Cases - Four-axis parallel review: Correctness, Standards, Spec, and Tests reviewers run concurrently, each reading the full touched files and callers rather than the diff in isolation, and write reports to agent/reviews/<range>/. - Smell baselines: Ships SMELLS.md (Fowler-style code smells) and TEST-SMELLS.md (tautological expectations, indiscriminate inputs, mocked collaborators, and more) so reviews have a fixed baseline even when the repo documents no standards. - Light mode for specless work: Unspecced changes get a single reviewer combining the Correctness and Standards briefs, with no Spec axis. - Use Case: Ask to "review this branch against main" before merging; the Skill pins the merge-base, spawns the axis reviewers, aggregates their reports, and disposes each finding as a fix commit, a filed ticket, or a declined note. ## Quick Start Ask the agent to review the changes on the current branch since main using the code-review skill.

Frequently Asked Questions about code-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review a git branch against main with an AI agent?

Invoke the code-review skill and name the fixed point, such as main or a commit SHA. It computes the three-dot diff from the merge-base, spawns parallel reviewers for correctness, standards, spec, and tests, then aggregates their reports into fixed, filed, or declined findings.

What does the code-review skill check in a diff?

It checks four axes: Correctness (broken callers, contracts, edge cases), Standards (repo conventions plus a Fowler-style smell baseline in SMELLS.md), Spec (missing requirements and scope creep), and Tests (test smells and whether tests would catch the change's absence).

Can I review changes that have no spec or ticket?

Yes, specless work runs in light mode: a single reviewer combines the Correctness and Standards briefs, reads the full commit messages for orientation, and writes one report. There is no Spec axis since there is nothing to check against.

What happens to the findings after a review?

Each finding gets one of three dispositions: Fixed via a follow-up commit carrying a Workflow-stage: review trailer, Filed as a proposed ticket, or Declined with a recorded reason. Reports live under agent/reviews/<range>/ and die with the worktree.

When should I use light mode instead of the full four-axis review?

Use light mode for specless work, small diffs, or when explicitly requested. The full review is warranted when a spec exists, the diff is large, or the change touches contracts other code depends on.